-
Notifications
You must be signed in to change notification settings - Fork 11
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
added option to use tensorflow.keras
#123
Conversation
Thanks for this one -- I don't think we should / need to invoke environment variables though, since it makes the internal behavior rather obscure and hard to predict. It'd be better if people just change their code to use the appropriate API for their use case. |
Yeah I realized that and overwrote those changes (I think you're looking at an old version of the PR, I force pushed and updated the description a few weeks back) Instead I just added |
Not sure how that old commit slipped by - you were right. Fixed now |
Just thought I'd circle around on this one. It's a small change that adds |
Any update on when this will be merged? |
That depends on Brian - he's a pretty busy person so I'm not expecting it soon. But seeing as this PR is still up to date with master, it should be safe for now. I was working on that package a while back and it seems that for some reason I didn't add this specific branch as an install dependency. In the most up to date version (0.1.1), it will install pumpp from this branch. I also cleaned a couple things up and modified the API and README a bit to make things a bit more organized and clearer. So use the readme examples to update existing code (main change is that |
I can probably get back to this project in the next couple of weeks. Got a few other things on my plate right now, but this PR generally looks good. I'll need to get my head back into pumpp for a minute to think it through, but it shouldn't be too long until we merge this. (Maybe a mid-june release?) |
This build is failing because of the following line, which wasn't tested in main 😱 : Lines 126 to 128 in 4009672
The problem is that in bumping up to tf2, we lose the import tensorflow.compat.v1 as tf1
tf1.placeholder(...) The problem then becomes that this is incompatible with eager execution mode. We can disable this using Additionally, we might want to rename the tf options to "tf1" and "tensorflow1" for the 0.6 release. |
What does this implement/fix? Explain your changes.
I added a method to build
tensorflow.keras
from a pumpp.